RDF Graph Stores

bootlessbwakInternet et le développement Web

12 nov. 2013 (il y a 7 années et 9 mois)

342 vue(s)

RDF Graph Stores
Christopher J. Matheus
Bell Labs / Alcatel-Lucent
Message
RDF Graph Stores
(aka “triple stores”)
store
RDF triples
(often with “context”)
and permit querying (and maybe updating)
via the
SPARQL
query language.
Many
commercial
and
free open source
implementations, becoming
faster
and
more scalable
and all
standards
-based.
•RDF Triples:
subject -predicate –object
RDF Basics
RDF and the Semantic Web
•Linked Data

use URIs for all resources

use HTTP for de-referencing

use RDF to describe resources

cross-reference data sets
•Metadata and Semantics
–Ontologies
–Inferencing
–Ontology Languages:

RDF Schema (
RDFS
)

The Web Ontology Language
OWL
•W3C query language for RDF
•sub-graph
matching
with variables
PREFIX rdf<http://www.w3.org/1999/02/22-rdf-syntax-ns#> ...
SELECT ?C ?X
FROM <http://example.org/companies>
WHERE { ?C ex:employees ?X .
?C rdf:rdf so:Corporation . }
---------------
?C ?X
ex:ALU “79799”
SPARQL
ex:employees
What’s exciting in SPARQL 1.1?
•Property paths
•Nested queries
•UPDATE
•HTTP Protocol
•Federated Queries
•Entailment Regimes
SPARQL Inferencing Notation
(SPIN)
•W3C Member Submission (Feb. 2011)
•SPARQL-based rule & constraint language
•Enables OWL 2 RL inferencing via SPARQL
•Components:
–RDF based syntax for SPARQL queries
–constraint language
–rule language
–SPARQL function definition mechanism
RDF Graph/Triple Stores
•Triple stores –
not!
Need at least quads

named graphs, time, provenance, access rights
•Standards
-based: RDF, SPARQL, RDFS/OWL
•Notable Examples:

OpenLink Software
Virtuoso

Franz
AllegroGraph

Apache
Fuseki
(Incubator)

Ontotext
OWLim

Systap
BigData

Garlik
4Store/5Store

Antonio Garrote
rdfstore.js

many more:
35 on
http://www.w3.org/wiki/SparqlImplementations
Which RDF Store is best?
Depends on what you need:
•Large data sets?
•Query performance?
•Quick, easy up and running?
•Production environment?
•Distributed, scalable?
•Rapid prototyping?
Large Data Sets:
AllegroGraph
•August 16
th
, 2011:
1 Trillion triples loaded and queried on
an Intel-based HPC cluster-in-the-cloud
provided by Stillwater SuperComputing
•Load:
–1,009,690,381,946 triples
–just over 338 hours
–average rate: 829,556 triples/second
Query Performance:
Virtuoso
•OpenLink Software’s RDF Store built on
top of their “Universal Server”(rdbms)
•DBpedia SPARQL Benchmark (ISWC 2011)
153,737,776 triples 27,665,352 resources 25 query templates
download, unzip, chmod, execute:
http://incubator.apache.org/jena/documentation/serving_data/
./fuseki-server --update --mem /ds
./s-put http://localhost:3030/ds/data default config.ttl
./s-get http://localhost:3030/ds/data default
./s-query --service http://localhost:3030/ds/query
'SELECT * {?s ?p ?o}’
./s-update --service http://localhost:3030/ds/update
'CLEAR DEFAULT’
Ease of Use:
Fuseki
Production Environment
•4Store/5Store
•AllegroGraph
•BigData
•BigOWLim
•Oracle
•Virtuoso
Distributed, Scalable?
•LOD2 Large Scale Experiments (in progress)
–Berlin SPARQL Benchmark (BSBM) 3.1
–SARA Lisa cluster with 64 nodes
–VectorWise server with 40 cores and 1TB
–Systems under evaluation:

BigData, BigOWLim, Jena TDB, Fuseki, Virtuoso,
4Store
Rapid prototyping: rdfstore.js
•open source Javascript RDF Store
•runs in node.js or browser
•supports most of SPARQL 1.1
•reasonable performance on LUBM
•experimental feature: watch triples
https://github.com/antoniogarrote/rdfstore-js
RDF vs. NoSQL
RDF Stores
_
-
all data as triples
+
standards-based:
+data model (RDF)
+query language (SPARQL)
+serializations (XML, etc)
+
formal semantics
+
standard tools
+frameworks and APIs
+editors
+reasoners
NoSQL Stores
.
+
not restricted to RDF
-
no standards
-
no formal semantics
-
store-specific tools
Key References
•W3C SPARQL Implementations
–http://www.w3.org/wiki/SparqlImplementations
•W3C Wiki: Large Triple Stores
–http://www.w3.org/wiki/LargeTripleStores
•W3C Wiki: RDF Store Benchmarking
–http://www.w3.org/wiki/RdfStoreBenchmarking
•W3C RDF Working Group Wiki

http://www.w3.org/2011/rdf-wg/
•W3C SPARQL 1.1 Draft Recommendation
–http://www.w3.org/TR/sparql11-query/
chris.matheus@alcatel-lucent.com